Census Principles of Industry and Product Classification, Manufacturing Industries

INDUSTRY AND PRODUCT CLASSIFICATION

of all of its separately controlled (through stock ownership or other-wise) subsidiaries. The definition of the firm used by the Bureau isessentially similar to that used by FTC and SEC.

Auxiliary units and central administrative offices. Associated withthe establishment concept is that of the auxiliary unit, which like thecaptive manufacturing plant, is a phenomenon chiefly of large busi-ness. These are elements of integration that tend to increase thetotal amount of production (and presumably profit) per unit offinal output. The problem of classifying auxiliary units occurs par-ticularly when such units are separately reported or separately lo-cated from "operating" establishments of the same firm. The maindistinctions between an "operating" establishment and an auxil-iary unit are: (i) an auxiliary unit is engaged in nonmanufacturingactivity to facilitate the principal activity of establishments of thesame firm; its existence depends on the establishments it serves; and(2) it is operated for the use of the firm's own establishments, thatis, it is not operated commercially for other business concerns orindividuals. The SIC manual does not provide specifically for Sep.arate auxiliary units; they are assigned to the industry of the estab-lishments served. For example, a warehouse serving "operating"manufacturing establishments of the same firm has been classifiedby most agencies as an auxiliary unit in the manufacturing division;the SIC manual provides an industry for public but not for "cap-tive" warehouses. On the other hand, a separate captive paper boxmanufacturing plant whose entire output is transferred to dressmanufacturing plants of the same firm for packaging the dresseswould be considered a manufacturing establishment, classified inthe paper box industry. The SIC manual recognizes manufacturingfor interplant transfer as a manufacturing activity.

The main types of auxiliary activities carried on for own use bythe manufacturing establishments of a firm are: force account con-struction; power generation; warehousing and storage; repair andmaintenance of own facilities and equipment; testing, research, anddevelopment work; buying operations; shipping and delivery opera-tions; and garage operation.

The central administrative office of multi-unit manufacturingfirms represents another type of unit in an establishment-reportingsystem. There are two schools of thought as to how these officesshould be treated for classification purposes. One says that theseunits are engaged primarily in administration and management ofthe firm, and the classification system should recognize this by creat-

23


ing a separate management industry. The other school claims thatthese units serve (administer, manage, keep records, etc.) otherestablishments of the same firm just as the executive, administrative,and clerical employees of a single-unit establishment serve otherparts of the establishment.

In the 1947 Census of Manufactures, the Bureau collected reportsfor central administrative offices but did not compile data for themalong with that for "operating" establishments. The Bureau didnot collect schedules for separate activities auxiliary to manufactur-ing (separate warehouses, garages, maintenance shops, etc.). It isexpected that data for both auxiliary units and central administra-tive offices will be accounted for in the 1954 census.

Data from firms. Agencies collecting financial data use the firmrather than the establishment unit in compiling statistics by in-dustry. The FTC and the SEC. in their quarterly financial reportseries, obtain reports from corporations. The classification unit maybe a single corporation (independent or subsidiary), or a parentcorporation consolidating data for parent and all subsidiaries in itsannual report to stockholders. The Bureau of Internal Revenuedata in the Statistics of Income series for corporations are basedlargely upon returns of separately incorporated entities. Consoli-dated returns for parent and subsidiary corporations are permittedin certain cases. They accounted for 6 per cent of net income of allreturns showing net income in 1947.6

THE INDUSTRY CLASSIFICATION

THE Standard Industrial Classification Manual (SIC), Volume i,Manufacturing Industries (1945), was used with some minor modi-fications in the 1947 Census of Manufactures and subsequent an-nual surveys of manufactures. A description of the scope and char-acteristics of manufacturing establishments is contained in theintroductions to the SIC manual and the Census of Manufacturesvolumes.

The manufacturing universe is divided into 21 major groups(designated by 2-digit codes), subdivided into about 150 industrygroups (s-digit codes) which are further divided into some 470 in-dustries (4-digit codes). Most of the industries are defined mainlyin terms of establishments primarily engaged in manufacturing aspecific product or group of products. In some instances, distinctionsare made in terms of operations (examples: stamping, forging,

6 Statistics of Income for 1947, Bureau of Internal Revenue, p. 2.24


foundry, and machine shop industries); some, in terms of process ofmanufacture or stage of production (examples: sweaters made inknitting mill vs. cut-and-sew plants from purchased knit fabric; cokemade in beehive vs. by-product ovens). Some industries are of aservice type primarily servicing other manufacturing establishments,usually on a contract basis on materials owned by others (examples:dyeing and finishing textiles; apparel contracting; galvanizing,electroplating, etc., metal products; printing trade services). Impor-tant instances where the same products may be assigned to two ormore industries on the basis of differing types or levels of productionoperations performed are listed in Appendix Table A-6.

APPLICATION OF THE INDUSTRY CLASSIFICATION

THE 1947 Census of Manufactures inquired about input (employ-ment, manhours, earnings, cost of materials, inventories, capital ex-penditures, etc.) and output of products. Except for a few dozenindustries, the product inquiries called for shipments (and some-times, production) of a preprinted and precoded list of products ofeach of the industries. Additional information (process of manu-facture, method of distribution, materials consumption, etc.) wasrequested, where necessary, to permit industrial classification of theestablishments.

While the SIC manual establishes a system for dividing the uni-verse of economic activities into smaller industry categories, it doesnot deal at length with specific application of the system in assigningindustry classifications to establishments. It does provide that theestablishment be assigned to an industry on the basis of its majoractivity, and the manual also states that ". . . in most cases, the in-dustry assignment is determined on the basis of the principal prod-uct made in the establishment, but in a number of instances othercriteria are used However, SIC does not prescribe the measur-ing rod (value of sales or receipts, value added, employment, etc.),nor any more specific method of applying the SIC system.

Theoretically, the major activity of an establishment would prob-ably best be measured in terms of income produced or value added,but such information is generally not reportable in the detail neces-sary for determining an industry classification. In practice, the Bu-reau and most other establishment-report collecting agencies meas-ure activity in terms of value of sales of products or receipts forservices.

7 Standard Industrial Classification Manual, as cited, p. .

25


In the 1947 census, an establishment was assigned to or classifiedin an industry generally on the basis of the principal products made.The products made by each establishment were grouped accordingto the industries to which they belong, and the group of productsaccounting for the largest part of the total value of shipments of theestablishment determined its industry classification. This group ofproducts is said to be the primary products of the establishment aswell as of the industry which it defines; all other products made byestablishments classified in the industry are referred to as secondaryproducts.

In some instances, a knowledge of the types and value of productsshipped was not sufficient for determining the proper industry classi-fication of an establishment. It was also necessary to know the proc-ess used in the manufacture of the products, or the materials used,or other characteristics of the operations of the establishment. Forexample, an establishment primarily engaged in producing insulatedcopper wire was classified in Industry 3392, Wire drawing, if itpurchased copper rods, drew the rods into wire, and insulated thewire, but classified in Industry 3631, Insulated wire and cable, if itpurchased copper wire and insulated it. In both cases, the majorproduct value of shipments would be insulated wire.

There are a number of limitations to the above method of assign-ing an industry classification to an establishment. In the first place,the assignment is based upon the plurality of a group of products(of an industry) having the greatest value. Thus an establishment

can be classified in Industry A even if only a small proportion, say,30 per cent, of its total shipments consists of products of Industry A,provided the value for that group exceeds the value for any othergroup of products in an industry classification. However, this situa-tion probably does not occur very frequently, as evidenced by the90 per cent average "homogeneity" (see above) for all manufactur-ing industries.

Further, the use of value of shipments as a measuring rod is some-times inadequate in approximating income produced or valueadded. This occurs particularly for establishments engaged in twoor more distinct activities crossing economic division lines (manu-facturing, wholesale trade, retail trade, services, etc.). In many ofthese cases, the use of value of shipments may lead to misleadingresults since a dollar's worth of manufactured product shipmentsmay not be equivalent in terms of income produced to a dollar'sworth of, say, wholesale trade sales. For example, an establishment

26


reporting the manufacture of motors valued at $i million, and alsothe purchase and resale at wholesale of hardware at $i. million,would be classified in wholesale trade on the basis of value of sales.However, on the basis of income produced, one dollar's worth ofmanufactured product sales is probably more nearly equivalent tofive or six dollars of wholesale sales, and therefore the establishmentwould be classified in manufacturing. Fortunately, the number ofthese borderline cases is undoubtedly limited to isolated establish-ments within certain industries and probably does not affect a sig-nificant portion of manufacturing establishments.

During the processing of the 1947 Census of Manufactures, a num-ber of difficulties were encountered in arriving at establishment in-dustry classifications. Many of the difficulties occurred in the border-land between the manufacturing and nonmanufacturing division—in the fringe industrial segments where establishments were engagedin manufacturing activities combined with nonmanufacturing ac-tivities or where a significant part of the output of manufacturedproducts was attributable to establishments in nonmanufacturingindustries. It was often a problem to distinguish manufacturingfrom nonmanufacturing classifications for establishments engaged insuch activities as manufacturing and wholesaling of meat and poul-try products; manufacturing dairy products and distributing fluidmilk; manufacturing and sale at retail or wholesale of bakery prod-ucts, confectionery, prepared feed, fertilizers, awnings, venetianblinds, window shades, etc; manufacturing millwork and lumberdistribution; manufacturing and repairing truck bodies; sheet metalwork and special trade contracting; fabricated structural steel pro-duction and general construction, etc.

The Bureau of the Census has been conducting annual surveys ofmanufactures on a sample basis since 1949. In many instances whereonly the current major activity is used as a basis for classification, asmall actual change in activity may be reflected in an exaggeratedchange in industrial classification. For example, in 1947 an estab-lishment with i,ooo employees shipping $io million worth of prod-ucts including $.i million worth of butter and million ofcheese would be assigned to the butter industry. If the butter andcheese figures were reversed in 1949, the establishment would beassigned to the cheese industry on a current activity basis. Thiswould shift the whole i,ooo employees, $io million shipments, etc.,to the cheese industry, whereas actually the change affected only$200,000 of shipments, and approximately 20 employees. To avoid

27


these abrupt and unrealistic fluctuations, the Bureau applied a "re-sistance" factor which would permit only significant changes inreal activity to be reflected in industry code changes. The resistancefactor was derived so as to minimize the combined errors of in-dustry trend and current level, assuming equal importance (weights)for trend and level.

The resistance factor technique was applied to a universe of ap-proximately 3o,ooo of the larger establishments in the 1949 samplesurvey of 45,000 establishments. Of these 30,000 establishments,primary activity changes occurred for i,ooo establishments. Applica-tion of the resistance factor prevented a change in code for 35 percent of these i,ooo cases. For the 15,000 smaller establishments, theold (1947) codes were maintained, that is, ioo per cent resistancewas applied.

As will be noted from the above discussion of the application ofthe industry classification, the nature and detail of informationavailable are important factors in determining industry classifica-tions of establishments. The detailed list of some 6,500 products andother inquiries on census schedules have resulted in establishmentclassifications often different from those assigned by other agencies.

For several years now the Bureau has been participating withother establishment-report collecting agencies (Bureau of Old-Ageand Survivors Insurance, BLS, Bureau of Employment Security),under the sponsorship of the Bureau of the Budget, in attemptingto achieve greater comparability of industrial statistics. An impor-tant beginning was the development of the SIC manual for use bygovernment agencies compiling data from establishment reports.However, a standard classification system is not enough, by itself,to accomplish the desired level of uniformity. Among other things,the system must be uniformly interpreted, the classification units—establishments—must be the same, and the same classification mustbe assigned to identical units for a given time period. The agenciesare at present directing their efforts toward these objectives.

3. Product ClassificationTHE 2 oo-odd schedules sent to manufacturers contained an aggre-gate of some 6,500 different products for which the Bureau of theCensus desired to collect shipments and, sometimes, production data.

A number of guiding principles were followed in the creation ofthe product list. An attempt was made to arrive at a balanced listof homogeneous products. In general, a product was not included if

28


its 1939 production was valued at less than $2 million. On the otherhand, product items which were reported in large volume by alarge number of establishments in 1939 were split up in 1947 intoa number of products, where feasible. In most cases, parts and ac-cessories of machinery and equipment were listed separately fromcomplete units. Distinctions were made in many instances betweenconsumer and producer goods; for example, household washingmachines, etc., were listed separately from commercial washing ma-chines, etc. Finally, each product was set up so as to be assignablein its entirety as "primary" to a particular industry; that is, onepart could not belong to one industry, and another part to anotherindustry. For example, dress gloves were divided between leatherand fabric gloves since establishments primarily making leathergloves were assignable to a different industry from those makingfabric gloves.

Basically, the products were organized or grouped into a structurerelated to origin of production, that is, according to the industryprimarily responsible for their output. This is in contrast to the1946 Standard Commodity Classification which follows the sequenceof the stage of production process. The latter system divides allproducts into three major groups: crude materials, fabricated basicmaterials, and end products. While the basic structures of the twosystems are organized differently, at the detailed product level, theyare comparable and consistent.

The 6,500 products were preprinted and precoded on the variousschedules so that it was generally necessary to assign codes onlywhen the respondent reported shipments of items not listed on hisschedule. Such items were written in on blank lines by the respond-ent and were assigned codes corresponding to one of the 6,500 prod-ucts of the system.

The term "products," as used in the Census of Manufactures, mayhave a broader or narrower content than in common usage. Forexample, automotive gasoline was reported as a single item. On theother hand, cotton broad-woven goods were distributed into nearly200 individual "products" according to type of weave, width offabric, and other specifications. For some items, e.g. bearings, itwould have been desirable to obtain product information in muchgreater detail than that actually requested, but the extent to whichthe production of individual types and sizes is concentrated in oneor two individual companies would have made it impossible to pub-lish detailed data. Thus the 6,500 individual products included on

29


the forms merely represent the number of items for which it wasconsidered practical to publish census information.

Of the 6,500 items included on the forms, data were actually pub-lished for approximately 6,oo. The balance were eliminated be-cause their publication would involve disclosure of the activities ofindividual companies or because a number of important producerscould not report products in the detail requested. A frequency dis-tribution of the value of product shipments is shown in Ap-pendix Table A-7.

To compile certain types of data, the 6,500 detailed productswere condensed into approximately i ,ooo broader classes of products.These product classes, with some modifications, were also used incollecting shipments data in the sample annual surveys of manu-factures for 1949, 1950, and 1951.

The extent to which industry and product statistics can be matchedwith each other is indicated in transition tables (Table s's, Censusof Manufactures, Volume ii, Statistics of Industry) which show, onthe one hand, the proportions by value of the primary and secondaryproducts shipped by the industry and, on the other, the value ofthe primary products of the industry made as secondary productsin other industries.

Following is an illustration of the relation between industry andproduct value data for the oleomargarine industry and the product,oleomargarine.

SHIPPED BYOleomargarine Other All

Industry Industries IndustriesPRODUCT (millions of dollars)

Total shipments ofoleomargarine industry 215

Oleomargarine (primary product) s 6 237Secondary products of

oleomargarine industry 42Salad dressings 36Shortening and salad oilsOther secondary products

The above table shows that establishments in the oleomargarineindustry shipped products valued at $215 million, only $i7 millionof which consisted of oleomargarine. On the other hand, total oleo-margarine shipments amounted to $237 million of which $64 mil-lion was contributed by industries other than the oleomargarineindustry.

In a classification system dividing the manufacturing field into

30


some 470 separate industries to which establishments are assignedon the basis of their principal activity, a certain amount of over-lapping production by one industry of products of other industriesis bound to arise. Such overlapping is particularly prevalent in suchindustry groups as apparel, furniture, and metal fabricating, whereestablishments within each group, employing the same basic typesof machinery and fabricating operations, produce a wide variety ofproducts belonging to a number of different industries. For 1947,the average amount of overlapping for all industries was approxi-mately io per cent in both directions, compared with from 15 to 20per cent for the furniture and metal fabricating (except transpor-tation equipment) groups of industries. The greater the degree ofoverlapping, the less meaningful are the relationships between gen-eral and product statistics for the particular industry.

31

SIC Industry or Industry Group201, Meat products202, Dairy products203, Canning and preserving204, Grain mill products

2051, Bakery products, except biscuit, etc.2052, Biscuit, crackers, and pretzels2085, Distilled, rectified, and blended liquors2111, Cigarettes2232, Woolen and worsted fabrics2271, 2273, Carpets and rugs2274, Hard-surface floor coverings, n.e.c.281, 282, Industrial chemicals283, Drugs and medicines

3o11, Rubber tires and tubes314, Footwear, except rubber321-323, Glass and glassware331, 832, Primary steel

3s34, Primary aluminum3411, Tin cans and tinware3431, Plumbers' supplies352, Agricultural machinery357, Office and store machines371, Motor vehicles and parts372, Aircraft and parts

Census of Manufactures: 1947(947 Value Percentage of Total

Added, Total Industry Shipmentsfor Industryb by First Four

(mill. $) Companies, 1947c$1,281 max.

595 max. 36917 max. 29

1,002 max. 281,101 i6

265 72472 75368 go6oo 28248 max. 49

83 8o2,006 max. 56

749 max. 6650 77786 max. 28713 max. 6o

3,780 max. 4065 100

232 78156754 max. 52504 max. 6t

,8sg max. 54955 max. 58


STATISTICAL APPENDIXTABLE A-i

Comparison of Concentration Ratios Derived from 1947 Census of Manufactures, EstablishmentReports and 1947 Federal Trade Commission Sample of Corporation Reportsa

FTC EstimatedPercentage

of TotalNet Capital

Assets Ownedby First Four

Companies, 19476g6o39363071858830

94523°88476255

1009674757471

44a Industries selected are s of the 26 industries shown in the Federal Trade Commission report.

Census data are based on classification of individual establishments. To determine concentrationratios, establishments of same company in a particular industry were consolidated; a companycould have establishments in many different industries. FTC data are based on classification ofcorporation as a whole (parent corporation and its subsidiaries), so that the corporation isclassified in only one industry.

b Value added data for all establishments of industry listed to indicate size of the variousindustries.

c Percentages based on value of shipments of largest 4 companies hi each SIC industry. Ratiosshown for combinations of 2 or more industries represent a maximum possible percentage thatwould be obtained if the first 4 companies of each of the industries of the combination wereidentical. Ratios for these combinations are prefixed with the symbol "max." (maximum). Ratiosare based on value of shipments, except for the following industry groups containing individualindustries with extensive duplication in shipments figures: 201, meat products; 2271 and 2273,carpets and rugs; and 332, primary steel; 371, motor vehicles and parts; 372, aircraft andparts. Ratios for these industries are based on "value added" data.

Sources: Census of Manufactures: 1947, Bureau of the Census, Vol. ii (value added data);Study of Monopoly Power, a report for H. Subcommittee prepared by the Bureau of the Censusfrom 1947 Census of Manufactures schedules, December i, 1949; and The Concentration of Pro-ductive Facilities, 1947, Federal Trade Commission, 1949.

32


TABLE A-2Frequency Distribution of Industries by Total Employment Size Class, 1947

Employment Size Class Number of Number of Value Added byof Industry Number of Establishments Employees Manufactures

(thousands of employees) Industries (thousands) (thousands) (mill. $)All industries, total 453 240.9 14,294 $74,426Less than i.o 9 .2 6 9

1.0— 1.9 24 i.6 34 2192.0— 2.9 19 1.8 46 3113.0— 3.9 20 2.1 68 3944.0-.. 4.9 19 2.3 84 4475.0— 9.9 95 17.6 709 3,942

10.0—24.9 ii8 40.0 s,86g 10,00925.0—49.9 70 50.9 2,521 13,63550.0—99.9 52 51.9 3,523 18,974

100.0 and over 27 72.5 5,434 26,456

Source: Census of Manufactures: 1947, Bureau of the Census.

TABLE A-3Frequency Distribution of Industries by Value Added Size Classes, '947

Value Added Size Class Number of Number of Value Added byof Industry Number of Establishments Employees Manufactures

(millions of dollars) Industries (thousands) (thousands) (mill. f)All industries, total 453 2409 14,294 $74,426Up to $4.9 13 .6 '2 46$ 5-4 9.9 25 i.8 45 199

10— 24.9 6o 8.4 249 1,09725— 49.9 86 16.9 700 3,29050— 99.9 83 21.7 1,162 6,090

100— 249.9 99 65.3 2,949 15,119250— 499.9 57 52.0 3,648 20,158500— 999.9 22 37.0 2,873 14,475

$i,ooo and over 8 37.2 2,656 13,952Source: Census of Manufactures: 1947, Bureau of the Census.

TABLE A.4Frequency Distribution of Industries by Degree of Homogeneity of Activity, 1947

Percentage Number of Number of Value Added byIndustry Homogeneity Number of Establishments Employees Manufactures

Class Intervala Industries (thousands) (thousands) (mill. $)All industries, total 453 240.9 14,294 574,426

50—59 1 .1 4 i86o-.6g 6 .9 50 43970—79 37 7.4 520 2,77280-89 129 60.6 3,767 19,23890—99 255 162.0 9,400 47,820

100 25 9.9 553 4,139

a 1947 value of shipments of primary products of each industry as a per cent of the industry'sshipments of all products. Average homogeneity for all industries is go per cent.

Source: Census of Manufactures: 1947, Bureau of the Census, Vol. ix, Statistics by Industry,Table 5'S.

33


TABLE A-5Frequency Distribution of Industries by Percentage Coverage of Primary Activity, 1947

Percentage Number of Number of Value Added byIndustry Coverage Number of Establishments Employees Manufactures

Class Intervala Industries (thousands) (thousands) (mill. )All industries, totalb 449 2385 14,257 $74,252Less than 50 19 9.5 366 1,737

50—59 9 3.9 86 4796o—6g 19 5.7 216 1,23170—79 50 18.2 838 4,2268o_89 110 6o.o ,ig5 1645890-99 207 1182 7,423 38,212

100 35 29.6 2,133 11,909a 1947 value of shipments of primary products by each industry as a per cent of shipments of

the products by all industries. Average coverage for all industries is approximately go per cent.b Excludes the following primarily service-type industries: 3465, enameling and lacquering;

3466, galvanizing; 3467, engraving on metal; and 3468, plating and polishing.Source: Census of Manufactures: 1947, Bureau of the Census, Vol. is, Statistics by Industry,

Table 5's.

TABLE A-6Principal Product Groups Primary to Two or More Manufacturing Industriesa

Product Group Industries in which Primary(A) (B)

Prepared meats 2011. Meat packing, wholesale2013. Sausages and other prepared meat products

Process cheese 2022. Natural cheese2025. Special dairy products

Flour, blended and prepared 2041. Flour and other grain-mill products2045. Blended and prepared flour

Sugar, refined 2062. Cane-sugar refining2063. Beet sugar

Knit apparel 2253-2255. Knit outerwear, underwear, and glovemills

23. Various apparel industriesWaterproof outer garments 2385. Raincoats and other waterproof outer gar-

ments3099. Rubber industries, not elsewhere classified

Other apparel and fabricated 23. Various apparel industriestextile products

Box shook 2421. Sawmills and planing mills, general2444. Wooden boxes (except cigar boxes)

Converted paper products 264-269. Paper product industries except pulpand paper board productsb

Soap 2841. Soap and glycerin2842. Cleaning and polishing preparations

Fertilizers 2871. Fertilizers, manufacturing and mixing2872. Fertilizers, mixing only

Coke 2931. Beehive coke ovens2932. By-product coke ovens

84


TABLE A-6 (Continued)

Product Group Industries in which Primary(A) (B)

Lubricating oils and greases 2911. Petroleum refining2992. Lubricating oils and greases not made in

petroleum refineriesLaminated glass 3211. Flat glass

3231. Glass products made of purchased glassGlassware, decorated 321. Flat glass

322. Pressed or blown glass and glassware3231. Glass products made of purchased glass

Ferroalloys and other 3311. Blast furnacesadditives 3313. Electrometallurgical products

Refined unalloyed nonferrous 333. Primary smelting and refining, nonferrousproducts metals

334. Secondary smelting and refining, nonfer-rous metals

Nails and spikes 3392. Wire drawing3481. Nails and spikes

Wire, except insulated 3312. Steelworks and rolling millsNonferrous rolling and drawing

3392. Wire drawingFabricated wire products, 3392. Wire drawing

except insulated wire 3489. Wirework, not elsewhere classifiedInsulated wire and cable 3312. Steel works and rolling mills

Nonferrous rolling and drawing3392. Wire drawing3631. Insulated wire and cable

Fire-control equipment 1941. Sighting and fire-control equipment3831. Optical instruments and lenses

a Establishments primarily engaged in manufacturing products listed in Col-umn A are classified in one of the two or more industries shown in Column B,on the basis of types of operations performed or materials used. For example,an establishment primarily manufacturing blended and prepared flour fromgrain milled at the same establishment is classified in 2041, Flour and othergrain-mill products; from purchased flour, in 2045, Blended and prepared flour.

b Most of the industries in major group 23, apparel and other finished productsmade from fabrics and similar materials and in industry groups 264-269, paperconverting industries, are defined in terms of establishments primarily engagedin making products from purchased materials (fabric for major group 23, andpaper and paperboard for industry groups 264-269). The presumption is that ifthe establishments make the same products from their own materials (producedat the same establishment) they are to be classified elsewhere (weaving mill orpulp and paper mill industries).

Source: Bureau of the Census.

85


TABLE A-7Frequency Distribution of Value of Individual Products Published in

Census of Manufactures: 1947Value of Shipments

Class Interval Number of Per Cent of Total(millions of dollars) Products Number of ProductsAll products, total 6,292 100

$ .0-4 1.95 1,539 242.0— 3.9 913 154.0— 5.9 630 106. 7.9 426 78. 9.9 384 6

10.0— 14.9 554 915.0— 19.9 361 620.0— 29.9 445 730.0— 39.9 251 440.0— 49.9 144 2

50.0— 59.9 8g 1

6o.o— 69.9 9070.0— 79.9 80 1

8o.o— 89.9 42 1

90.0— 99.9 38 1

100.0—199.9 170 3200.0299.9 54 1

300.0—399.9 33 1

400.0—499.9 14 X

500.0—599.9 6 x600.0—699.9 4 X

700.0—799.9 6 x800.0—899.9 3 X

900.0—999.9 5 X

$i,ooo.oandover 11 x

a A large proportion of the products of small value are of the unavoidable"residual" type or of the type needed to simplify classifications or instructions forreporting other products.

x Means less than .5 per cent,Source: Census of Manufactures: 1947, Bureau of the Census.

COMMENTSOLOMON FABRICANT, National Bureau of Economic Research

ECONOMISTS belong to that hungry tribe of whom it is said thatwhen shown a finger they try to seize the hand. Those of us con-cerned with business concentration do not admit the richness ofthe census for our purposes; we point only to the obvious fact thatthe census can be made richer—with no additional cost as far aswe are concerned.

We ask, for example, for information on net capital assets, whichwould provide us with a more stable measure of size than does out-.

36


put, value of product, or other flow items. It is true that some diffi-culties were encountered when the possibility of securing data oncapital assets was explored in a pre-test of the 1947 census. How-ever, if the census can secure useful information on capital expendi-tures, it is difficult to believe that it cannot at the same time secureuseful information on capital assets.

More information on central bffices and related activities wouldalso be useful. Many of these offices straddle a variety of industries,and we need to know more about their relative importance beforewe can be sure we are measuring individual industries adequately.We need to know more also about ownership connections betweenindustries, and not only within manufacturing but also betweenmanufacturing and mining, trade, etc.; and central office informa-tion would help. A "distribution of sales" schedule also would help.

At present we are in some degree victims of the way in which thework of the Bureau of the Census is divided between its IndustryDivision and its Business Division—not to speak of its AgricultureDivision. Ownership connections are broken when an enterprise issplit between the Census of Manufactures and the Census of Trade.But even for enterprises entirely within manufacturing, our meas-ures of business concentration depend on the classification of estab-lishments followed by the Bureau of the Census. It would be in-formative, therefore, if the Bureau told us ho its classificationscompare with those implicit in various administrative rulings, suchas those made by the Wage and Hour Division. And we ought tobe given a clearer idea as to how congruent the census industriesare with the clusters of establishments organized in trade associa-tions and similar groups.l

Besides extending the census to provide additional data, the Bu-reau can produce a valuable body of source material merely bymaking new arrangements of data now in its files. Generally speak-ing, this means providing breakdowns and cross classifications ofvarious sorts. Aggregates are only the beginning of information.

For example, we would like to know how individual establish-ments or enterprises change in size from one year or census to an-

1 When discussing the classification question, Conklin and Goldstein mentiona "resistance" factor to avoid "abrupt and unrealistic fluctuations" in the scopeof industries. It would be interesting to know in which industries the resistancefactor is most important. It would be well also if the Bureau of the Census in-dicated whether it expects to adhere to resistance factors even when a new censusis taken, for apparently the factor has been applied so far only in moving for-ward from the 1947 census via the annual surveys.

37


other. Light on this question would be provided if for each industrywe had a cross tabulation of establishments, by size in one yearagainst size in another year, size being measured by value added,employment, or value of product, or—if the information were ob-tained—capital assets. It might not be necessary to do this for allestablishments; a sample in each industry might be sufficient. If wehad such information, we would know the extent to which estab-lishments or enterprises shift into and out of the "top four," andcould deal with one of the questions that Rosenbluth raises in hisdiscussion of the time period. More generally, information would beprovided about the shifts that occur in the position of individualfirms wherever they may be in the size distribution. We might learnsomething also about the movement of establishments between in-dustries; or within industries, of shifts in major product or extentof specialization.

Everybody knows the story of concentration in livestock purchas-ing. We could do with a few fresh examples of concentration inbuying. What we need is a classification of a wide range of industrieswith respect to concentration in buying of various materials. I sus-pect that the census already has a fair amount of the basic data; per-haps all that is needed is to tabulate them in suitable form.

Suits mentions the relationship between product and industry.Tabulations already available in the Census of Manufactures pro-vide information for narrow groups of industries. It would be de-sirable if, on occasion, the census could publish larger segments ofthe full product-by-industry tabulation that covers all industriesand products.

Information already available could be organized to indicate howconcentration in production of end products in a given industry isrelated to concentration in production of intermediate materials inthat industry. We would like to know also how concentration in pro-duction is correlated with concentration of labor in trade unions inthe same industry. Available information could probably be or-ganized to show how frequently oligopolies face one another intheir transactions.

Apart from the question of developing and extending our data,there is the problem of using the data most effectively to measurebusiness concentration.

Value added or net value added has been suggested as superiorto value of product in measuring concentration because materialsand fuel differ in importance among establishments. However, even

38


value added is not free of difficulties associated with varying degreesof integration. Companies that do their own construction and main-tenance work or other auxiliary activities will necessarily have alarger value added than will concerns that do not, yet this kind ofintegration is largely irrelevant to our purpose. A further point:value added includes profits. When profits are high, value addedwill be high and so will profit rates. If we take value added as ameasure of size and correlate it with the rate of profit, as we some-times do, we may get spurious results. This is one of the reasons forasking that the census try to obtain information on capital assets.

A problem arises when using value of product in concentrationmeasures because "captive" establishments transfer their output toaffiliated establishments at assigned values. These values are usuallylower than market values, as may be seen in the 1947 census sta-tistics on the blast furnace industry, and the concentration ratiosmay be too low also. In any case, the presence of these "arbitrary"values—to use the term favored by Conklin and Goldstein—is dis-turbing. Perhaps better than value of product, then, is physical vol-ume. One could omit "captive" establishments, as Conklin and Gold-stein suggest; but what is really needed is knowledge of their im-portance, and of the extent to which shifts occur in the proportionbetween transferred and sold outputs.

It may be noted, further, that in some industries substantialquantities of goods are produced and consumed in the same indus-try. These goods will not find their value reflected in the total valueof the output of that class of goods, since the total value will usuallyrelate to the quantities sold.

Many of our concentration measures are based on industry classesas defined in the census and similar sources. Yet there may well beconsiderable competition between, say, manufacturing industries onthe one hand and nonmanufacturing industries on the other. Thusthe manufacture of canned fruits and vegetables is closely competi-tive not only with the manufacture of dried fruits and vegetablesbut also with the sale of fresh fruits and vegetables and with homecanning. Another and perhaps better example is cheese or butter,which are (or used to be) produced on farms as well as in factories.

Whatever we do, difficulties will be encountered in getting upsensible concentration measures and using them. It must be em-phasized, however, that these difficulties are not merely technicalmatters. They reflect phenomena of the economic world that arethemselves worthy of study by economists. We need to know, for

39


example, when auxiliary activities are taken on, when they aresloughed off. We need in fact a theory of auxiliary activity_recallStigler's recent paper on specialization.

Another illustration is found in the type of analysis performedby Rosenbluth in which he distinguishes between changes in con-centration within industries and changes in the weights of differentindustries and asks what has been the effect of each type of changeon the over-all concentration ratio. A real question here is whetheror not there is an economic relationship between these two "inde-pendent" factors. More specifically, is there any tendency for highlyconcentrated industries to grow less rapidly because they are morehighly concentrated? A related question looks the other way alongthe line of causation. How does stage of growth of an industry in-fluence the degree of concentration that characterizes it? Further,how does degree of concentration fluctuate with the business cycle?These questions, raised by Moses Abramovitz in an article nowfifteen years old, still need looking into.

The stage we have reached in our study of concentration in busi-ness is much like the stage reached in the study of concentration ofincomes a decade or so ago. At that time we thought the majorquestion was the shape of the income distribution. We have sincegraduated to a higher level in which we worry about the factorsthat determine the position of a family in the income distributionand the bearing of that position on the division of income betweensavings and consumption. We seem to be approaching this newlevel of analysis in the present field. If we learn from the experienceof investigators of income distribution, we should rise to the newlevel sooner.

FRANK J. KOTFKE, Federal Trade Commission

THE construction of a concentration ratio poses a problem in classi-fication. Use of a single concentration ratio involves acceptance oftwo categories as comparable and relevant. Contrast of two or moreconcentration ratios generally predicates their comparability. Thepossibilities are great that the most readily available data are notcomparable, or, though comparable, are not relevant to the problemat hand. Consequently, economists must take care to avoid the misuse of data in studies of economic concentration.

Publications of the Industry Division of the Bureau of the Censusare the principal basis for concentration ratios for manufacturing

40


activities in the United States. Conklin and Goldstein have renderedan important service in explaining precisely the classification prin-ciples on which such data are developed, and in indicating some ofthe considerations to be reviewed before any figure is incorporatedin a concentration ratio. Many matters are involved in defining thecategories of a census study. The Bureau of the Census must con-sider not only the interest of students of business concentration, butalso the needs of other scientists, lawmakers, businessmen, and laborunions. The Bureau must consider the bases upon which inform-ants can supply data, and also the importance of establishing speci-fications that all informants will interpret in the same 'way. Finally,the Bureau endeavors to accommodate its reporting program tothose of other government statistical agencies. Even if there wereonly one concept of economic concentration, adjustment to theseseveral needs almost certainly would leave much of the data lessthan ideal for the measurement of concentration.

A few comments on two of the industries listed in Appendix Aillustrate difficulties for which a student of concentration must beon guard. Consider the dairy-products industry in 1947: The Fed-eral Trade Commission (FTC) estimated that the four leading com-panies accounted for 6o per cent of the net capital assets, while theBureau of the Census estimated that the four leading companiesaccounted for not more than 36 per cent of the value of shipmentsand interplant transfers originating in this industry.' Conklin andGoldstein imply that the difference is attributable to the procedureof the FTC in this particular study in assigning to one industry theentire capital assets of each parent company. Yet well over 90 percent of the capital assets of the four leading dairy companies weredirectly related to the manufacture and distribution of dairy prod-ucts. National Dairies had a minor commitment in frozen foods andsalad products. Borden had seven chemical plants and eight special-products plants, but for the products of many of these plants, milkwas an important raw material—viz., casein, milk sugar, beveragebases, and prescription foods. Carnation made its own cans andoperated several feed mills. Another circumstance that might con-ceivably account for such a discrepancy is that the leading com-panies held large capital assets outside the continental UnitedNora: The views expressed in this Comment are the writer's and not necessarilythose of the Federal Trade Commission.

I The report of the Federal Trade Commission was issued August 24, 1949. Thedata on which the census ratios are based were released December 1, 1949.

41


States. But in 1951 (the closest year for which the information isavailable), less than 2 per cent of the assets of National Dairies waslocated abroad. Borden in 1947 operated in two Canadian prov-inces, and two of Carnation's thirty milk plants were in Canada:this would account for a difference of not more than a few percent.age points in the two concentration ratios. The denominator forthe FTC ratio, of course, included the value of nondairy assets, andof capital assets outside the continental United States, for all com-panies primarily engaged in manufacturing dairy products.

The difference between the FTC ratio of 6o per cent and themaximum ratio of Conklin and Goldstein is traceable to aspects ofthe dairy-products business not covered by the census data theyemployed. All establishments distributing fresh milk and cream,including those establishments for which the manufacture of dairyproducts was the more important part of their operations, wereexcluded in developing the census data.2 Establishments for receiv-ing milk, establishments for storing dairy products, and facilitiesfor transporting dairy products all were excluded. The result is aconcentration ratio based on a much more limited conception ofthe dairy-products industry than that which the FTC had in mind.

As another example, newspaper and business writers constantlyrefer to the "automobile industry." However defined, this industryprovides employment for the largest group of manufacturing work-ers in the United States. The following are some of the conceptsthat may be considered relevant to a study of concentration in theautomobile industry (with concentration measured, for convenience,in terms of the four leading companies):

i. The percentage accounted for by the four largest producers ofthe value (f.o.b. plant) of all passenger automobiles shipped duringa year. A close approximation, for which data are available, is theper cent of new passenger-car registrations. The four largest com-panies accounted for 88 per cent in 1947.8

2. The percentage accounted for by the four largest producers of2 In 1947, combination mercantile and manufacturing plants manufactured

$1.7 billion of dairy products, whereas exclusively manufacturing plants ac-counted for $.6 billion. In 1948, sales of milk dealers and dairy-product storeswere $2.9 billion.

8 Registration data are from Automotive News, 1952 Almanac Issue, p. 32. TheFTC suggested that 1948 production figures were preferable to those for 1947, asstrikes, material shortages, and other factors seriously disturbed production in1947. However, census data are available for 1947 but not 1948; consequently,Conklin and Goldstein had no choice but to use 1947.

42


the value (f.o.b. plant) of all motor vehicles shipped during a year.The term "motor vehicles" as used here includes passenger auto-mobiles, ambulances, buses, and trucks, but not road-building equip-ment, self-propelled construction equipment, combat vehicles, farmmachinery, locomotives, or motorcycles. The best approximationavailable is the per cent of new passenger-car and truck registrations.The four largest companies accounted for 85 per cent of such regis-trations in 1947.

. The percentage accounted for by the four largest producers ofthe value of shipments (including miscellaneous receipts) of allestablishments in which motor vehicle production is of greater valuethan output of goods primary to any other industry, as defined bythe Bureau. The four leading companies here are those whose motor-vehicle establishments so defined, had larger shipments than did theestablishments in the motor-vehicle industry controlled by any othercombination of four companies. As of 1935 their share was 87 percent. (Subsequently, the Bureau of the Census abandoned this con-cept of the industry.)

4. The percentage accounted for by the four largest companies ofthe value added by manufacture by establishments in which pro-duction of motor vehicles, motor-vehicle parts, or both motor ve-hicles and motor-vehicle parts is of greater value than output ofgoods primary to any other industry, as defined by the Bureau.Here the four companies are those whose motor-vehicle establish-ments so defined, added a larger value by manufacture than did theestablishments in the motor-vehicle and parts industry controlled byany other combination of four companies. This is the measure usedby Conklin and Goldstein. The share of the first four companies in1947 was 56 per cent.

5. The percentage accounted for by the four largest companies ofthe net capital assets of all companies in which capital resourcesdirected to the production of motor vehicles, motor-vehicle parts,or both motor vehicles and motor-vehicle parts are of greater valuethan those committed to any other industry as defined in sufficientlybroad terms to facilitate classification of parent corporations in asingle industry. With this approach, the leading companies are thosehaving the largest net capital assets. This is the concept used by theFTC in its report on concentration. The share of the four largestcompanies in 1947 was estimated at 71 per cent.

6. The percentage accounted for by the four largest companies of4 The Concentration of Productive Facilities, i7, FTC. 1949.

43


the total assets of all companies in which capital resources directedto the production of motor vehicles, motor-vehicle parts, or bothmotor vehicles and motor-vehicle parts are of greater value thanthose committed to any other industry, as defined in sufficientlybroad terms to facilitate classification of parent corporations in asingle industry. Here the leading companies are those with thelargest total assets. On this basis the share of the four largest com-panies at the close of 1947 was 68 per cent.

Concept 1 is, of course, a "product" concept. For purposes ofmarket analysis, some economists might prefer to divide it intothree or four price classes. Its purpose, and that of 2 and 3 as well,is to measure the fraction of the supply of a commodity providedby the largest companies. Concept 2 recognizes that automobiles,buses, and trucks generally are produced and distributed by thesame companies, often with the same facilities. Some persons mightconsider this an industry concept, although in Bureau of the Cen-sus terminology it is a combination of certain "product classes."

Concept 3 corresponds to the census basis for measuring 1947concentration in 440 of the 452 industries recognized in tabulationsfor that year. In measuring concentration, it serves as a substitutefor 2. Sometimes it is a poor substitute, for reasons described byConklin and Goldstein. However, the Bureau of the Census nolonger recognizes a motor-vehicle .industry, presumably because im-portant integrated producers cannot develop satisfactory estimatesof their production of motor-vehicle parts. For the motor-vehicleand parts industry, which it does recognize, value added by manu-facture is used instead of value of shipments plus interplant trans-fers, since the motor-parts fraction of the latter would be countedagain in the value of the assembled motor vehicles. This is con-cept 4. It gives weight to'integration within a limited area of themotor vehicles industry, although it is not a complete measure ofthe extent to which motor-vehicle producers are integrated, even inthat segment of their business classified as manufacturing.

Concept 5 is a measure of "total economic strength or productivepotential." So also is concept 6, which is identical in coveragebut employs a different criterion of size.6 In this they should be dis-

5lbid., p. 12.6 M. A. Adelman, among others, has urged total assets as preferable to net capi-

tal assets. The issues are developed in "The Measurement of Industrial Con-centration," Review of Economics and Statistics, November 1g51; John M. Blair,'The Measurement of Industrial Concentration: A Reply," and M. A. Adelman,"Rejoinder," both in Review of Economics and Statistics, November 1952.

44


tinguished sharply from concepts i, 2, and 3, which are measures ofthe share of the supply actually accounted for by the largest pro-ducers. According to the reasoning of the FTC in presenting thisreport, a very large company may affect competition in its majorline of activity not only through its current participation but alsobecause of the other resources upon which it may draw. In manyinstances a very large company could transfer establishments cur-rently engaged in other lines to the production of its principal prod-uct. Such reserve capacity is a significant circumstance, accordingto some students of oligopo1y. Establishments not susceptible tosuch a transfer generally bear a "vertical" relationship to the manu-facture of the company's principal product: they supply materialsand components, perform additional processing for certain enduses,, distribute or service the product, or perhaps utilize by-productsof the main operation. Ownership of such establishments may be asource of strength to a large company in its major line of activity.8

The Commission did not extend the analysis to the petroleum in-dustry or to other industries where "capital assets . . . significant inrelation to the size of the corporation [were considered to] con-tribute only indirectly to the corporation's position in the industry,and are not convertible to the industry in which the corporation isclassified." Also excluded were industries where one of the leading

"A large corporation's productive equipment which is engaged in turningout products in 'other' industries may be convertible to the industry in whichthe corporation is classified. As was so strikingly illustrated by the experienceof World War II, a very large proportion of modern technology is highly flexibleand convertible. Hence, from the point of view of measuring a large corporation'sproductive potential, it might be unrealistic to exclude from the industry inwhich the corporation is classified that portion of its equipment which hap-pens at the moment to be engaged in an 'outside' industry, but which couldhe quickly converted to the industry in which the corporation is classified." TheConcentratiun of Productive Facilities, 5947, as cited, p. io.

8 The Commission cited as an example the ore boats of a large metal refiner,ibid., pp. g-io. A more recent report of the FTC staff describes the importanceof vertical integration in certain situations: Monopolistic Practices and SmallBusiness, 1952, pp. 21, g.

9 Ibid., p. '1. Ratios on the primary steel industry were published only aftercomparison with industry data on capacity suggested that such limitations wouldnot result in any serious overstatement of concentration. The Commission alsorelied on a special tabulation of the returns of the fifty largest manufacturingcompanies in the Census of 1937. As summarized by the staff of the TemporaryNational Economic Committee (TNEC), this tabulation (which suppressed theidentity of companies and industries) showed that "the major portion of thetotal value of products of those companies was accounted for by the value con-tribution of relatively few products." Willard L. Thorp and Walter F. Crowder,The Structure of Industry, TNEC Monograph 27, 1941, Part vi, p. flog.

45


companies was classified in another industry, and industries where"specialization by one leading company on certain products, . . . bya second leader on other products, and so on, was so extreme thatthe figures for the industry as a whole were not significant."0 Theratios published were "no more than estimates. . . . Errors, however,probably do not exceed a few percentage points

Thus for all economists who have occasion to construct measuresof concentration, or to use measures developed by others, the Conk-lin-Goldstein paper is both an admonition and an aid. It is a warn-ing that not all concentration ratios can be interpreted in the sameway; it is a warning also that some data can contribute little to themeasurement of concentration. The paper is valuable as a referenceon the use of the Bureau's data on manufactures and on the inter-pretation of measures of concentration developed from such data.

But it seems to me that this informative paper holds lessons forproducers of data as well as for consumers of data. For example,the comments on beet sugar and cane sugar and on tin cans andglass containers point up our lack of a description of the participa-tion of an enterprise across many industries. Tabulations are de-sirable that would indicate the extent to which substitute commodi-ties are (and are not) produced by the same companies, the pro-portion of the supply of a raw material that is captive to specificconsuming industries, and the proportion of still other industriescontrolled by the companies producing their principal raw ma-terials. Within recent months an encouraging beginning has beenmade on this problem by the Bureau of the Census in cooperationwith the FTC. A still wider description of enterprises is desirablethan can be afforded by a study of manufacturing alone. The con-current censuses of the mineral industries, manufacturing, and tradein 1954 provide the opportunity for valuable tabulations. It is tobe hoped that the Bureau is not denied the resources to derive fromthese censuses a description of the major structural features of oureconomy. Business decisions are made by firms, and the Bureau'spublications will be made much more useful by explicitly recogniz-ing that firms do not confine themselves to a single establishment, orto establishments in a single industry.

There is ample reason for revising many heterogeneous cate-gories—both industries and product groups. Of the general charac-teristics of an industry named by Conklin and Goldstein—competi-

10 The Concentration of Productive Facilities, '947, as cited, p. '4.'1 Ibid., p. '5.

46


tive nature of products, competitive nature of firms, similarity ofmethod of manufacture, similarity of facilities, "other physical ortechnological factors" (among them, presumably, significant sizeand sufficient enterprises to permit publication of data), and his-torical precedent_there are numerous cases in which historicalprecedent and a generalized complementarity of products seem theprincipal elements of cohesion. This observation deals only inci-dentally with the n.e.c., or "not elsewhere classified," genre. Suchcategories are inescapable, and cause difficulties only when one ormore types of establishment in a given residual assume the impor-tance of a separate industry. Food preparation, not elsewhere classi-fied, with 1951 shipments exceeding $2 billion, is overdue forattention.12 So also are four other n.e.c. industries with shipmentsexceeding $i billion, and nine with shipments of less than $i billionbut more than $'/2 billion.

Given the industry concept, the problem that led to the "resist-ance factor technique" is inevitable once sampling is substituted forcomplete enumeration. This technique is the practice described byConklin and Goldstein of not reclassifying an establishment to re-flect a change in its primary activity where the discrepancy therebyintroduced in the current level of the industry is smaller than thediscrepancy avoided in the industry trend.13 The fatal weakness ofthe present technique is that the entire body of industry data de-veloped by a survey is incapable of interpretation. A better policywould be to return the industry to the Standard Industrial Classi-fication concept, and to introduce a new concept, the "historicalindustry," in which all establishments would retain the same classi-fication as in the last census year. An alternative is to abandon thepublication of the survey data on industries appreciably affected byshifts in the primary activity of establishments, and to place greateremphasis on obtaining product st4tistics, which are not affected bythe "resistance" problem.

The greater success of the 1951 and 1952 surveys in developing sta-tistics on product classes than the 1950 or 1949 survey is most en-couraging. Yet even for 1952, data are lacking for half the classes,

12 About two-thirds of the output of this "industry" is roasted coffee. Otherimportant products are peanut butter, ground spices, ready-to-mix desserts,potato chips, and sweetening syrup and molasses.

13 Establishments with less than 100 employees have been continued in their1947 industry without review. While there is little likelihood of such establish-ments shifting into or out of most industries, dairy and apparel industries wouldseem to be important exceptions.

47


and for many of the categories on which figures are supplied thestandard errors are so large that the data are practically useless. Inthis situation the data on product class groups14 are most welcome,but even these are not available for a third of the groups, andamong those for which data are available the estimates frequentlyare subject to large standard errors.15 Product statistics—detailed,current, and reasonably accurate—are important in so many waysthat the Bureau of the Census should have vigorous support in itsefforts to improve this phase of the annual survey of manufacture.

DANIEL B. SUITS, University of Michigan

Tur proper evaluation of classifications can only be made in termsof the objective to be achieved by the use of the resulting classes,and different objectives generally require different classifications.These comments are restricted to classifications suitable for themeasurement of concentration on the supply side of the market.

There are a number of attributes or yard sticks—capital, employ-ment, sales, value added, production, etc_by which we can, atleast ideally, obtain a magnitude for each of a set of economic units(firms, establishments) and a total for the set. This total can thenbe compared with the total for any given subset of the units (thelargest one, the four largest, etc.) or the magnitudes of the indi-vidual units may be somehow distributed and analyzed. But thesemeasures and the concentration analysis based on them are not, soto speak, meaningful in their own right. They are, rather, indirectways to approach the measure of something else, which we maycall economic power over a market. They derive whatever validity

14 A product class group' covers all products primary to a given industry.Thus, conceptually, there is a product class group corresponding to each industryrecognized in census statistics. In the margarine example presented by Conklinand Goldstein, $237 million is the i947 value of shipments of the product classgroup margarine, whereas $215 million is the value of shipments of the marga-rine industry. In this instance, margarine is the only product class in the group,but commonly there are several classes. For example, the product class group"canned and preserved products, except fish and meat" has nine product classes:canned fruits; canned vegetables and specialties; canned fruit juices; cannedvegetable juices; canned baby foods; canned soups and poultry products; jams,jellies and preserves; and bulk fruit and vegetable juices.

15 For product classes, 158 of the 633 estimates are subject to a standard errorof from 6 to io per cent of the estimate, and s i8 estimates are subject to a standarderror of 15 per cent, For product class groups, estimates are available for 154 cate-gories where a group consists of more than one product class. Of these, 48 aresubject to a standard error of 6 to io per cent, and i6 are subject to a standarderror of 15 per cent.

48


they have from the extent to which they succeed in representing agiven output market completely and to the exclusion of other out-put markets.

If each firm or establishment commonly exercised influence inonly one market, the problems of defining product market classesand industrial classes would be the same. Unfortunately under anyreasonable definition many establishments—not to say firms—par-ticipate in a number of different markets. Thus the appropriateclassification of firms or establishments into industries involves asomewhat greater measure of adjustment and compromise than theclassification of products into markets, which in turn does not lenditself to the ideal results we might prefer to obtain.

The appropriate classification of products is based on the com-petitive relations which hold among them. These competitive rela-tions are not fixed, but change over time with alteration of produc-tive techniques, consumer tastes, the introduction of new products,and doubtless in some cases the general level of business activity.Moreover, the competitive relations among products and firms differin the long run, in which ultimately almost any firm is a competitorof any other, and the short run in which productive facilities areby and large fixed. It is the latter on which any given census focusesattention; thus our objective is a classification that will group to.gether the products of rival productive facilities.

The commodity is merely the physical "embodiment" of theservice of the establishment. The economic power possessed by anestablishment inheres in the service it can provide, and to measurethis service purely on the basis of the particular application beingmade of it at a given moment would be clearly wrong. Thus closetechnical substitutes, which are literally alternative embodimentsof the same service, should be classified in the same product market.For example, a particular die-casting establishment that producesrefrigerator door handles at one time and automobile radiator orna-ments at another is part of a market in which similar facilities com-pete with each other over a wide range of uses. An appropriateclassification of products must recognize these alternatives and iden-tify them with a single market class of products.

There are, of course, all degrees of technical versatility and somerecognition must be given this fact. Ideally we can conceive of usingsome minimum level of the elasticity of technical substitution as acriterion. Unfortunately, the basis on which to estimate these elas-ticities is seldom available and we are restricted to more readily ob-

49


servable characteristics of products. Two products can be consideredtechnical substitutes if they are commonly produced in substantialamounts in the same establishment, by essentially the same equip-ment, technical process, and labor, and among which frequent varia-tions in the proportion of output are observable. The existingcensus emphasis on the technical structure of production is a reflec-tion of this criterion. Likewise, of course, it is the fact of technicalsubstitutability that justifies our disposition to consider productsof differing sizes and specifications as constituting a single marketclass. On the other hand we would consider as distinct two productsfound in the same establishments but produced by distinct tech.nical processes. Variation among them represents alternative usesof working capital, but not of the fixed facilities themselves. Norwould joint products, appearing in effectively fixed proportions, gen-erally be considered the same product. The establishments compete inthe production of the output combinations, but frequently each prod-uct enters competition with others that are not subject to the sametechnical restriction. Cottonseed oil and cottonseed cake and meal, forexample, each have important competitors and are therefore distinct.Where the several joint products do not have important "outside"rivals, however, they need not be separated.

The second way in which services substitute for one another is,of course, through the production of substitute products. Marketclasses should not distinguish among the production of differentcommodities that are close substitutes to the consumer. Ideally theappropriate measure here would be the cross.elasticity of demandsamong products, but again we must have reference to more readilyobservable criteria. Those commodities serving generally similarpurposes, and among which users are frequently observed to varythe proportions of their purchases in response to price variations,may be defined as consumer substitutes.

Broad classes of products may be set apart "by eye," but finaldetermination requires intimate technical knowledge of the struc-ture of production, the nature of products and their uses, andfamiliarity with the habits of their consumers. It would obviouslybe of advantage if less subjective criteria could be applied, and sub-stitutes identified by the behavior of their prices, sales, margins,etc., as actually observed in the market. The difficulties, however,are considerable, as the following example suggests.

The fact that the prices of close consumer substitutes shouldmove together might serve as a necessary criterion for classification.

50

TABLE 1Variation in Price Ratios, Selected Pairs of Commodities, igi-ig

Coefficient of Variation ofLog of Price Ratio

(Per Cent)Commodity PairPackers' steer hides, nativePackers' steer hides, TexasMen's tan dress welt shoes, calf leathersMen's tan dress welt shoes, side leatherElectrolytic copper ingot#8 bare copper wireAmerican medium saltGranulated bulk salt6-foot crosscut sawsGranulated bulk saltOleomargarine, white, ChicagoButter, extra firsts, ChicagoBone blackLamp blackCurrantsRaisinsOne-horse walking plowsSouthern single warp cotton yarnPackers' steer hides, TexasAmerican medium salt

a

Source: Prices are Bureau of Labor Statistics wholesale prices.

native steer hides, and calf-leather and side-leather men's shoes—are, of course, close consumer substitutes, but are also closely re-lated in production. They are followed by the production-related,

1 The coefficient of variation is the standard deviation measured as a per centof the mean. Thus it is unaffected by absolute levels. Using the log of the priceratios frees the measurement of the objection that price ratios are necessarilybounded at zero, but have no upper bound.

51


It is obviously not sufficient since prices may be kept together bysubstantial common cost elements, or even move together quite for-tuitously. The absolute levels of most prices are highly correlatedover wide cyclical changes, but the behavior of the price ratios canbe analyzed.

In order to obtain a measure of price variation that is unaffectedby the absolute levels of the prices and that also gives the same re-suit no matter which price is selected as the denominator of theratio, the coefficient of variation of the log of the price ratio is taken.1

In Table i, ten commodity pairs are ranked in order of the sta-bility of their price ratios. The two top ranking pairs—Texas and

5.7

1.75

4.2

6.s

8.4

9.0

11 0

12.1

12.9

21.2


nonsubsti tu te pair, copper ingot_copper wire. Oleomargarine—but-ter shows slightly more price variability than bulk salt—crosscutsaws, while bone black—lamp black and currant—raisins show onlyslightly less variability than one-horse plows—cotton yarn. A muchgreater number of price ratios would need to be examined beforethe merit of the method could be assessed, but it is clear that thebehavior of quantities, production costs, and other data must beincorporated in the analysis before price behavior can be adequatelytested.

Ideally we might hope to apply the criterion of substitutabilityto partition the universe of commodities into mutually exclusiveclasses so that (i) any pair of close substitutes fall in the same classand, (2) any two commodities in the same class are a pair of closesubstitutes. The first objective can always be achieved. It would,indeed, be satisfied by summarily lumping all goods in a singleclass. Moreover, forming the maximum number of product classesfor which (i) will hold true will carry us a considerable way towardthe attainment of (2). But the complete satisfaction of the secondobjective is not necessarily possible.

The difficulty is easily explained in terms of the logic of relations.In order to satisfy both (1) and (2) the relation "close substitutefor" must, among other things be transitive.2 That is if A and B andB and C constitute two pairs of close substitutes, it must followthat A and C are close substitutes.

Even among consumer substitutes, this transitivity does not al-ways hold. Commodities are often arranged in "chains" where eachis a close substitute for its immediate neighbors, but the nearnessof substitution becomes less as we compare a given good with thosefarther away.8

When we have both consumer and producer substitutes, the diffi-culty is even greater. If A is a close technical substitute for B, whileB is a close consumer substitute for C, A and C need bear no directsubstitute relationship to each other, while the relationship be-tween A and some fourth good, a close technical substitute for C,may be even more tenuous.

The magnitude of this problem should not be exaggerated. The2 The substitute relation must also be reflexive (any good is a close substitute

for itself) and symmetrical (if A is a good substitute for B, then B is a goodsubstitute for A). The three properties taken together mean that the relation'close substitute for" is an equivalence relation, i.e. that it will produce preciselythe partitioning of the universe of commodities that we want.

S An obvious example of this in another context arises in spatial competition.

52


merit of discussing the underlying logic of classification derives fromthe light shed on the process of formulating and evaluating themost useful classes for the purpose at hand. If strict application ofthe criterion of substitution yields classes that are too heterogeneous,we can frequently subdivide them. This will put some substitutecommodities in separate classes, and what is gained by increasedhomogeneity must be balanced against this loss. Wherever a rela-tively small amount of production serves as a bridge between twolarge and otherwise unrelated classes we are clearly justified in sub-dividing the classes. Where this is not the case, the balance of ad-vantage probably lies with retaining the larger class as a whole.

The decision is again based on subjective considerations, but wecan push them back a stage by making a class subject to subdivisionif it contains a product which (i) if completely ignored would re-suit in an increase in the number of market classes and (2) if itswhole production were assigned to either of the resulting classesthe measure of output concentration employed would not be sub-stantially affected.

Aside from the considerable technical knowledge required, thereis no reason why such a system of product classes could not be builtup from census product detail. Census product classes on the mostdetailed level (e.g., bottled soft drinks, carbonated, containing kolaextract; emulsified asphalt paving materials; builders' door locks,lock sets, and lock trim; etc.) are quite clearly suitable elements forsuch classification. In fact many of the larger census "bold facetotal" classes (e.g., soft drinks, total; paving mixtures and blocks,total; etc.) are themselves internally homogeneous with respect tosubstitution. On the other hand, even at the "bold face total" levelthere are classes which are not. (Professional furniture, total, in-cludes both hospital bed springs and laboratory cabinets and cases,which are not close substitutes in either sense.) In some cases eventhose product classes that explicitly define census industries are ele-mentary in this sense, but in general the system of classes must bebuilt up from product classifications below the census industry level.

Where economic concentration is to be related to other variablesobtainable only on a firm or establishment basis, the economic unitsthemselves must be classified into industries which follow as closelyas possible the lines of the product classes. This is done by assign-ing each establishment or firm to the industry in which the greatestportion of its output falls.

It is evident that industries based on market classes defined as53


above will differ somewhat from the existing census classification. Inparticular, establishments separated by the census solely on the basisof the material or technical process employed would appear in asingle class. Moreover, although census emphasis on technical simi-larity tends to cause establishments in a single census industry to falltogether in the "new" classification, doubtless some of them wouldbe pulled apart. Even here, however, we should probably find thatin most cases the readjustment of industry boundaries required toobtain adequate levels of homogeneity and coverage would resultin their recombination.

The problems surrounding homogeneity and coverage arise pri-marily out of the nature of the firm's activities themselves and arenot materially altered by changing the system of classification em-ployed. An acceptable industry must be one for which both homoge-neity and coverage, measured in terms of the products that definethem, are high. Given fixed product classes, homogeneity and cov-erage tend to behave inversely. We can always increase the coverageof an industry by transferring into it establishments originally classi-fied in other industries, but we generally do so at the expense ofhomogeneity. Where raw industrial classes must be adjusted, somecompromise between the measures must be accepted.

It is this fact that would probably tend to recombine census in-dustries which the strict application of product classification wouldsplit. A group of establishments, most of which were engaged insome production of each of two nonsubstitute commodities, wouldbe split between the two classes involved. If, however, there are noother important substitutes for these commodities, combining thetwo raw classes would yield an industry of high coverage in termsof either product, while the homogeneity of the combination withrespect to either product would be only slightly lower than whentaken over the separate raw groups. As a matter of fact, I suspectthat we would be disposed in such case to question the usefulnessof the product distinction itself and combine the two sets of prod-ucts into a single class.

In any case, the areas in which meaningful industry classificationsare difficult or impossible to attain are precisely those in whichexisting census classifications are least meaningful, and for exactlythe same reason. These are areas in which integrated and noninte-grated production are both common. Classification on the basis ofsubstitutability of product only formally removes the distinctionbetween integrated production and production carried on with pur-

54


chased materials. Semifinished and finished products are substituteitems of sale to the establishment producing both, but they are inno sense close substitutes in production. The integrated establish-ment clearly belongs in two distinct industries and whether classi-fication is by the finished or the semifinished product, the result isunacceptable. We can gain high coverage or homogeneity in oneindustry only at the expense of a low measure in the other, andcombining both reduces the homogeneity measure well below whatcould be obtained from either.

Ultimately, getting usable industry classifications in these areasdepends on our ability to redefine establishments and obtain mean-ingful data from their separate departments. Where this cannot bedone, the only alternative is to restrict our analysis of concentrationin these areas to product data.

55

Date post:	03-Feb-2022
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

Census Principles of Industry and Product Classification, Manufacturing Industries

Documents